Ecosystem monitoring is central to effective management, where rapid reporting is essential to provide timely advice. While digital imagery has greatly improved the speed of underwater data collection for monitoring benthic communities, image analysis remains a bottleneck in reporting observations. In recent years, a rapid evolution of artificial intelligence in image recognition has been evident in its broad applications in modern society, offering new opportunities for increasing the capabilities of coral reef monitoring. Here, we evaluated the performance of Deep Learning Convolutional Neural Networks for automated image analysis, using a global coral reef monitoring dataset. The study demonstrates the advantages of automated image analysis for coral reef monitoring in terms of error and repeatability of benthic abundance estimations, as well as cost and benefit. We found unbiased and high agreement between expert and automated observations (97%). Repeated surveys and comparisons against existing monitoring programs also show that automated estimation of benthic composition is equally robust in detecting change and ensuring the continuity of existing monitoring data. Using this automated approach, data analysis and reporting can be accelerated by at least 200x and at a fraction of the cost (1%). Combining commonly used underwater imagery in monitoring with automated image annotation can dramatically improve how we measure and monitor coral reefs worldwide, particularly in terms of allocating limited resources, rapid reporting and data integration within and across management areas.

Karimunjawa National Park is one of Indonesia’s oldest established marine parks. Coral reefs across the park are being impacted by fishing, tourism and declining water quality (local stressors), as well as climate change (global pressures). In this study, we apply a multivariate statistical model to detailed benthic ecological datasets collected across Karimunjawa’s coral reefs, to explore drivers of community change at the park level. Eighteen sites were surveyed in 2014 and 2018, before and after the 2016 global mass coral bleaching event. Analyses revealed that average coral cover declined slightly from 29.2 ± 0.12% (Standard Deviation, SD) to 26.3 ± 0.10% SD, with bleaching driving declines in most corals. Management zone was unrelated to coral decline, but shifts from massive morphologies toward more complex foliose and branching corals were apparent across all zones, reflecting a park-wide reduction in damaging fishing practises. A doubling of sponges and associated declines in massive corals could not be related to bleaching, suggesting another driver, likely declining water quality associated with tourism and mariculture. Further investigation of this potentially emerging threat is needed. Monitoring and management of water quality across Karimunjawa may be critical to improving resilience of reef communities to future coral bleaching.

Coral reefs are declining rapidly in response to unprecedented rates of environmental change. Rapid and scalable measurements of how benthic ecosystems are responding to these changes are critically important. Recent technological developments associated with the XL Catlin Seaview Survey have begun to provide high‐resolution 1‐m² photographic quadrats of ~1.8–2 km of coral reefs at 10 m depth by using a semi‐autonomous image collection SeaView II camera system (SVII). The rapid collection of images by SVII can result in images being taken at a variable distance (1–2 m) from the substrate as well as having natural light variability between images captured. This variability can affect the quality of taxonomic resolution archived from photographic quadrats captured by SVII. Conventional approaches for taking photographic quadrats of coral reefs involve taking images from a fixed distance with the use of artificial light. These methods often provide images that enable high taxonomic resolution, but are typically used to cover areas of 50–150 m. Here, we select key metrics associated with coral reef condition from a functional perspective to contrast how much is lost in terms of association and agreement from image annotations using SVII images (1‐m² photoquadrats) compared to conventional methods, capturing 0.5‐m² photoquadrats using a digital single lens reflex camera from a fixed height of 0.5 m above the benthos. Comparisons were made over the same 50‐m linear transects at 15 sites in the Maldives. Photoquadrats were manually analyzed using an online image annotation tool and image repository called coral net with 25 randomly placed dots per 0.5 m² of a quadrat. Our results reveal high levels of correlation and agreement between methods when measuring the abundance of hard corals as a functional group and individual labels, which make up hard corals as a functional group. These results demonstrate the two methods are comparable when measuring functional groups and coral communities on coral reefs. Therefore, involving rapid semi‐automated technologies that maximize data collection for monitoring coral reefs does not necessarily imply that taxonomic resolution is compromised. This insight has important ramifications for detecting important changes in coral reef condition, such as a decline in coral cover or shifts in benthic community composition.